A novel systematic approach for robust numerical simulation of carbon fiber-reinforced plastic circular tubes: Utilizing machine-learning techniques for calibration and validation

Abstract

Developing a reliable and robust finite element model of a carbon fiber-reinforced plastic (CFRP) composite structure is investigated by using the LS-DYNA solver and Python. This study tries to provide a systematic numerical approach to cover the principal impediment to adaptation of composite energy absorbers, that is the lack of a reliable predictive method. The proposed procedure aims to further the understanding of advanced composite structures’ behavior during the crash phenomenon by developing an accurate finite element model. To do so, the mechanical properties of the material were extracted from American Society for Testing and Materials (ASTM) standard test methods, followed by experimental investigation of circular CFRP tubes undergoing quasi-static loading. A numerical simulation framework was then utilized to scrutinize the effectiveness of simulation parameters on the crushing mechanism. Finally, a systematic approach based on machine learning techniques was performed to adjust non-physical modeling parameters for further calibration and validation. In this regard, a versatile Python code was developed to automate pre-processing, processing, and post-processing steps. The code also provides a groundwork to perform machine learning techniques. Interestingly, the numerical and experimental results were highly correlated with a correlation coefficient of almost 90%. Additionally, several non-physical numerical parameters were found to be inactive, while some else were identified as effective parameters, and their corresponding effectiveness was quantitatively extracted and discussed for the first time in the literature.